Case–Control Study of Risk Factors for Hospitalization Caused by Pandemic (H1N1) 2009

TOC summary: Pregnancy, asthma, diabetes, and a history of smoking were among risk factors found.

The fi rst cases of pandemic (H1N1) 2009 infection from New South Wales (NSW) were reported in May 2009. In NSW, laboratories were required to notify all pandemic (H1N1) 2009 diagnoses to the NSW Department of Health under Public Health Act 1991 (21). Australian public health management protocols recommended laboratory testing for all persons with infl uenza-like illness (fever and cough or sore throat) admitted to a hospital (22). Public health follow-up was required to ascertain hospitalization status for all notifi ed cases at the time of diagnosis; this information was collated on a statewide database. Within Sydney, the capital city of NSW (population 4.4 million), there are 4 Area Health Services (AHSs) responsible for the provision of local public health and clinical services.
By the end of June 2009 (before a vaccine was available), community transmission was widespread in Australia, and public health efforts were focused on protecting those at greatest risk for severe disease. The groups considered most vulnerable were pregnant women; indigenous people; very obese persons; persons with preexisting chronic medical conditions, including lung, heart, and kidney disease; and persons with blood, metabolic or neurologic disorders, immunosuppressive conditions, or asthma (23). Persons in these risk groups and those with severe disease were urged to seek medical attention early if infl uenza-like symptoms appeared. Doctors were provided with free antiviral medication for patients who were seen within 48 hours of symptom onset. A vaccine became publicly available for distribution in September 2009. To help reach populations who would most benefi t from prevention and early intervention, we sought to identify independent risk factors for moderate to severe disease from pandemic (H1N1) 2009 infection among adults and to describe the characteristics of those who sought early medical treatment to determine the effectiveness of public health messages.

Methods
The study population was defi ned as persons >16 years of age residing in metropolitan Sydney during July 1-August 31, 2009. All interviews were conducted during September and October 2009.

Cases
We defi ned a case as a person with infl uenza-like illness admitted (for a minimum of an overnight stay) to a Sydney metropolitan hospital from July 1 through August 31, 2009, who had laboratory confi rmation of pandemic (H1N1) 2009 by PCR that was notifi ed to the Department of Health. Patients <16 years of age or residing outside metropolitan Sydney were excluded from the study because we assumed the threshold for hospital admission may have been different for this age group and for patients in regional and rural areas. To ensure complete case ascertainment, all cases notifi ed in the study period were cross-matched with the Department of Health database containing all hospital admissions in NSW for the study period; additional identifi ed cases were included in the study. If case-patients were not able to complete a telephone interview because of ill health or disability, another household member completed the interview on their behalf. Up to 30 attempts were made to contact case-patients by telephone.

Case-Control Study
In a case-control study, we compared demographic and health information reported by case-patients with those of controls. We defi ned a control as a person >16 years of age residing in metropolitan Sydney who had not been hospitalized for infl uenza in 2009. Telephone numbers (stratifi ed by AHS) were used to randomly select potential control households. Within each AHS in Sydney, we selected 2 households per case. A single control was randomly selected from within each selected household for interview. Up to 12 attempts were made to contact selected households. Fewer attempts were made to contact controls than case-patients as a result of random-digit generation methods described in detail elsewhere (24). The age and sex distribution of participating controls was compared with that of the metropolitan Sydney population.

Study Questionnaire
We used a standard questionnaire to ask case-patients and controls about infl uenza symptoms, pregnancy or delivery within the previous 28 days, weight and height, smoking history, current and previous medications, and past hospitalizations. In addition, information was collected regarding diagnosis of general health conditions, including asthma, lung disease (defi ned as emphysema, chronic lung problems and/or chronic bronchitis), heart disease (defi ned as heart problems from birth, rheumatic heart disease, angina, heart attack, and/or heart failure excluding hypertension), diabetes (type I or II), other metabolic disorders, kidney disease (defi ned as kidney transplant, renal failure, and/or dialysis), liver disease, blood disorders (defi ned as sickle cell disease, thalassemia, or hemoglobin problems), mental health diagnoses, neurologic conditions (defi ned as conditions that involve muscles, nerves, or the brain), immune suppression (defi ned as cancer, HIV infection, or immunosuppressive medication), and obstructive sleep apnea. Ethics approval was not required because information was collected under NSW Public Health Act 1991 (21)

Statistical Analyses
In univariate analysis, the proportions of characteristics among case-patients and controls were compared by using χ 2 tests. In multivariate analysis, independent risk factors for hospitalization were assessed through logistic regression by using backward elimination. All variables with a univariate level of signifi cance p<0.25 were selected for inclusion in the base model, and variables were excluded if the p value was >0.05 and did not meaningfully alter the point estimates of the remaining variables. Because a similar proportion of case-patients and controls reported a diagnosis of asthma not requiring regular medication, only a history of asthma requiring regular medication was included in the fi nal logistic regression model.
Additional logistic regression models were constructed to compare case-patients with controls for women of childbearing age (16-45 years), case-patients who received mechanical ventilation compared with all controls, casepatients who sought medical attention within 48 hours of onset of symptoms compared with case-patients who sought medical attention after 48 hours, and controls who reported symptoms of infl uenza-like illness during the study period (defi ned as self-reported fever and either cough or sore throat) with controls who reported no illness during the study period. Mechanical ventilation (rather than intensive care unit admission) was used as a measure of severity of illness because cases were reported from several hospital facilities with varying criteria for patient admission to intensive care. The model fi t the data well by the Hosmer-Lemeshow goodness-of-fi t test (χ 2 = 15.85, 9 df, p = 0.07). Seventynine percent of pairs were concordant, and c = 0.804.

Results
In total, 402 hospitalized patients were identifi ed as eligible for inclusion in the study. Of these, 302 (75%) participated in the study, 27 (7%) refused interview, 66 (16%) were unable to be contacted, and 7 (2%) were excluded because of language diffi culties. In univariate analysis, there was no signifi cant difference between participating patients and nonparticipating patients with respect to sex (p = 0.226), geographic location of residence (p = 0.341), indigenous status (p = 0.123), length of stay in hospital (p = 0.477), or ventilation status (p = 0.890). However, interviewed patients were signifi cantly younger (median 45 years, range 16-88 years) than patients who were not interviewed (median 51 years, range 17-88 years) (p = 0.007).
None of the case-patients reported giving birth in 28 days before symptom onset. Of the 40 patients who were pregnant at the time of symptom onset, 28 (70%) were in their third trimester. Sixty-six (22%) case-patients were current smokers, and 91 (30%) were ex-smokers. Among the 91 ex-smokers, the median time smoked was 20 years (range 1-60 years, median 15 cigarettes/day); 39 reported cessation >5 years prior to illness, 15 patients between 12 months and 5 years, and 37 patients within the past 12 months. Among the 66 current smokers, the median time smoked was 20 years (range 1-60 years; median 10 cigarettes/day) (online Appendix Table).

Case-Control Study
There were no signifi cant differences in characteristics of case-patients and controls by place of residence, receipt of 2009 seasonal infl uenza vaccination, or history of neurologic disorders (online Appendix Table). In univariate analysis, compared with controls, case-patients were more likely to be male, aged 16-35 years and 46-55 years, have higher body mass index (BMI), and report a history of asthma, heart disease, mental health diagnosis, immune suppression, obstructive sleep apnea, lung disease, diabetes, liver disease, blood disorder, and pregnancy, and smoking (online Appendix Table).
In the logistic regression model, age, sex, asthma (requiring regular medication), smoking (current or exsmoker), heart disease, immune suppression, lung disease, diabetes, and pregnancy were independently associated with hospitalization for pandemic (H1N1) 2009 (Table 1). Similar results were found when analysis was restricted to men only or women only. The factor most strongly associated with hospitalization was pregnancy, followed by lung disease, immune suppression, and ages 16-25 and 46-55 years (Table 1). Overall, 262 (86%) case-patients reported >1 independent risk factor (asthma, heart disease, immune suppression, lung disease, diabetes, pregnancy, or smoking) compared with 315 (52%) of controls (p<0.0001). The risk for hospitalization increased with increasing number of reported signifi cant risk factors.
In the logistic regression model for women of childbearing age, asthma (requiring regular medication), lung disease, diabetes, and pregnancy were independently associated with hospitalization for pandemic (H1N1) 2009 (Table 2). In total, 88% of women of childbearing age In univariate analysis, compared with controls, casepatients who required mechanical ventilation were more likely to report a history of lung disease, asthma (requiring regular medication), have a higher BMI, be pregnant, have an infl uenza vaccination in the previous 12 months, and be 26-45 years of age (Table 3). In the logistic regression model for ventilated case-patients, a history of lung disease, diabetes, pregnancy, high BMI, or status as a current or exsmoker were independently associated with mechanical ventilation (Table 3).
Information on the time from onset of illness to medical attention was available for 295 (98%) case-patients. Of these, 238 (81%) reported having >1 risk factor listed in the public message campaigns. Overall, the proportion of case-patients seeking medical attention was similar for both those with reported risk factors and those with no risk factors (80% and 79%, respectively). There was no signifi cant difference in individual underlying risk factors between case-patients who sought medical attention within 48 hours of symptoms and those who did not (Table 4).

Discussion
We found that pregnancy, lung disease, immune suppression, asthma, diabetes, heart disease, and a history of smoking were associated with hospitalization from pandemic (H1N1) 2009 infection. Among women of childbearing age, pregnancy was the single greatest risk factor for hospitalization, followed by diabetes, history of asthma requiring regular medication, and lung disease. Obesity was not an independent risk factor for hospitalization although it was a risk factor for mechanical ventilation. The majority of case-patients sought medical attention within 48 hours. This study did not identify any particular risk groups that were less likely to seek early medical attention.
Our study was designed to identify risk factors for moderate to severe illness resulting from infl uenza (as measured by requirement for hospital admission), not risk factors for acquiring infl uenza. Controls for the study were therefore selected from the community rather than nonhospitalized case-patients. When controls who reported infl uenza-like-illness were compared with controls without these symptoms, there was no signifi cant difference in risk factors other than age. This fi nding suggests that apart from age (which is likely to refl ect past infection with infl uenza strains that protected against pandemic [H1N1] 2009) (25), other participant characteristics were not important for determining susceptibility for infection. Our data are subject to several limitations. First, compared with the adult population in Sydney, controls were older and a higher proportion were women, introducing the possibility of bias if the groups were not otherwise similar. However, our analysis adjusted for age and sex, and the fi ndings were consistent with those of similar studies examining risk factors for seasonal infl uenza (3)(4)(5)6). When the model was restricted to gender, the results were similar to the fi nal model. These fi ndings suggest that the potential bias from control selection on the fi nal model was minimal. Second, risk factor status was determined by self-report for case-patients and controls. However, public messaging during the pandemic relied on  (26)(27)(28), kidney disease, neurologic disease, immune suppression, lung disease, asthma, smoking, and relatively young age (29,30) as the most common concurrent conditions for hospitalized patients (11)(12)(13)(14)(15)(16)(17)(18)(19)(20). Given that many of these underlying medical conditions do not occur in isolation, our analytical study was able to ascertain which of these were independently associated with hospitalization from pandemic (H1N1) 2009.
Although high rates of pandemic (H1N1) 2009 infection have been reported from indigenous people in other reports (31), our study lacked suffi cient power to explore the independent impact of being Aboriginal on the risk factor of moderate to severe disease. However, among those Aboriginal case-patients included, all reported a history of other independent risk factors. These data may suggest that it is the high prevalence of risk factors for severe disease that place Aboriginal people at an increased risk rather than genetic susceptibility. BMI was not independently associated with increased risk for hospitalization with pandemic (H1N1) 2009. Obesity appears to be a confounder of other risk factors for overweight patients. Of the 58 case-patients with BMI >35 (very obese), 55 (95%) reported >1 signifi cant risk factor, including smoking (35/55, 64%), asthma (32/55, 58%), and diabetes (18/55, 33%). Further analysis suggested that obesity was independently associated with increased risk of ventilation in our study. Of the 35 ventilated case-patients, 11 reported a BMI >35, and all but 2 patients reported other signifi cant risk factors.
Our study highlights the increased risk of moderate to severe illness from pandemic (H1N1) 2009 for pregnant women and introduces smoking as an independent risk factor for hospitalization from pandemic (H1N1) 2009. In addition, our study provides evidence to support the continuation of infl uenza prevention efforts (including vaccination) targeted to persons with lung disease, immune suppression, asthma, diabetes, and heart disease. Although Aboriginal status and obesity may not be independent risk factors for severe disease, they indicate the likely presence of other risk factors, and so prevention messages should continue to be directed to these groups.